Bug 11492

Summary:	mgaapplet dies without a trace (segfault on handling SIG_CHLD)
Product:	Mageia	Reporter:	Luca Olivetti <luca>
Component:	RPM Packages	Assignee:	Thierry Vignaud <thierry.vignaud>
Status:	RESOLVED OLD	QA Contact:
Severity:	normal
Priority:	Normal	CC:	mageia, nic
Version:	3	Keywords:	NEEDINFO
Target Milestone:	---
Hardware:	x86_64
OS:	Linux
Whiteboard:
Source RPM:	mgaonline-2.81-1.mga3.src.rpm	CVE:
Status comment:
Bug Depends on:	10289
Bug Blocks:

Description Luca Olivetti 2013-10-18 11:41:52 CEST

From time to time, mgaapplet dies without any message (i.e., the last message I see in syslog is "Checking... Updates are available").
I usually realize when I see there are updates in the rss feed but I'm not prompted by mgaapplet.
I saw other reports with "mgaapplet segfaulted" but all of them say that the bug has been caught by Drakbug and that doesn't happen here.
My desktop is KDE and I use NetworkManager (with the kde plasma applet) instead of mageia network center.


Reproducible: 

Steps to Reproduce:

Comment 1 Luca Olivetti 2013-10-18 12:17:21 CEST

It's a segfault, and it doesn't always happen always. Running it under gdb, this is a backtrace when it happens (it doesn't seem very useful to me, maybe it is to the developers):


(gdb) bt
#0  Perl_csighandler (sig=17, sip=<optimized out>, uap=<optimized out>) at mg.c:1381
#1  <signal handler called>
#2  0x00007ffff69ffd9d in poll () at ../sysdeps/unix/syscall-template.S:81
#3  0x00007ffff367a9a4 in g_main_context_poll (n_fds=3, fds=0x7fffe00010c0, timeout=-1, context=0xd4d670, priority=<optimized out>) at gmain.c:3584
#4  g_main_context_iterate (context=0xd4d670, block=block@entry=1, dispatch=dispatch@entry=1, self=<optimized out>) at gmain.c:3285
#5  0x00007ffff367ae02 in g_main_loop_run (loop=0xd4d600) at gmain.c:3484
#6  0x00007fffea0a9c56 in gdbus_shared_thread_func (user_data=0xd4d640) at gdbusprivate.c:277
#7  0x00007ffff369e235 in g_thread_proxy (data=0x1731770) at gthread.c:797
#8  0x00007ffff6cd3d18 in start_thread (arg=0x7fffe85c4700) at pthread_create.c:308
#9  0x00007ffff6a097cd in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:110
#10 0x0000000000000000 in ?? ()

David Walser 2013-10-18 14:52:37 CEST

Assignee: bugsquad => thierry.vignaud

Comment 2 Thierry Vignaud 2014-01-08 04:56:17 CET

I think I just fixed that one on Cauldron as part of bug #10289

I understand you can reproduce this one from times to times?
If yes could you try the fix from http://gitweb.mageia.org/software/drakx-net/commit/bin/net_applet?id=571aeab77c9d52375d061f53a729b3910f40e876 ?

Thierry Vignaud 2014-01-08 04:56:24 CET

Keywords: (none) => NEEDINFO

Thierry Vignaud 2014-01-08 05:00:29 CET

Depends on: (none) => 10289
Summary: mgaapplet dies without a trace => mgaapplet dies without a trace (segfault on handling SIG_CHLD)

Comment 3 Thierry Vignaud 2014-01-08 05:02:28 CET

@Colin: the bug you reported by mail is likely this one and was seldom seen on mga3.

CC: (none) => mageia

Comment 4 Luca Olivetti 2014-01-08 18:05:22 CET

I applied the fix but I have to wait for some updates to try and reproduce it.

Comment 5 Mageia Robot 2014-01-24 06:22:35 CET

commit 8e6a3f840cf1e9141942e93c757e1bc21732c91e
Author: Thierry Vignaud <thierry.vignaud@...>
Date:   Wed Jan 8 04:42:00 2014 +0100

    delay setting SIG_CHLD handler (mga#11492)
    
    thus fixing segfault on startup
    
    Conflicts:
    	NEWS
---
 Commit Link:
   http://gitweb.mageia.org/software/mgaonline/commit/?id=8e6a3f840cf1e9141942e93c757e1bc21732c91e

Comment 6 Mageia Robot 2014-01-24 06:22:38 CET

commit 7597a033b03c5e86531ce59ee44f714dbae07018
Author: Thierry Vignaud <thierry.vignaud@...>
Date:   Wed Jan 22 21:07:47 2014 +0100

    block CHLD signals on startup (mga#11492)
    
    in order to prevent glib-threading segfaults (mga#11492), just block the
    CHLD signal during the window where glib create threads behind our back
    (RT-120951)
    
    Conflicts:
    	NEWS
---
 Commit Link:
   http://gitweb.mageia.org/software/mgaonline/commit/?id=7597a033b03c5e86531ce59ee44f714dbae07018

Comment 7 Thierry Vignaud 2014-01-24 06:27:36 CET

BTW, You can test mgaonline-2.81.1 from core/update_testing once it lands on your favorite mirror
(you need to enable Core Update Testing in "Software Media Manager" in Mageia Control Center)

Comment 8 Luca Olivetti 2014-01-24 09:04:30 CET

After the fix in comment #5 mgaapplet spontaneously died at least once (in fact I found it strange that it modified just net_applet, which I don't use)
I manually applied the fix in comment #6

Comment 9 Thierry Vignaud 2014-01-24 18:20:20 CET

There's nothing about net_applet in comment #5
But you can just try mgaonline-2.81.1 from core/update_testing

Comment 10 Luca Olivetti 2014-01-24 18:39:40 CET

Sorry, it was comment #2

Comment 11 Luca Olivetti 2014-02-06 19:34:49 CET

Today I found that mgaapplet (with the patches in comment #5 and comment #6) was not running.

Comment 12 Nic Baxter 2015-03-31 03:49:29 CEST

Still current?

CC: (none) => nic

Comment 13 Luca Olivetti 2015-03-31 09:04:30 CEST

In the last few months I didn't see mgaapplet die, but I cannot tell if it's just luck or one of the latest version fixed it for good. I'm currently on mgaonline 3.10.1-1.mga4 build on the 22nd of february.

Comment 14 Thierry Vignaud 2015-03-31 09:06:46 CEST

Let's close as OLD for now then

Status: NEW => RESOLVED
Resolution: (none) => OLD